基于决策树方法的罗布泊“大耳朵”干盐湖区盐壳分类研究  

A Study on Salt Crust Classification Based on Decision Tree Technology in Lop Nur "Great Ear" Dry Salt Lake Area

作　　者：颜安[1] 蒋平安[1] 武红旗[1] 马黎春[2] 

机构地区：[1]新疆农业大学草业与环境科学学院,乌鲁木齐830052 [2]中国农业大学资源与环境学院,北京100094

出　　处：《遥感信息》2009年第2期57-63,75,共8页Remote Sensing Information

基　　金：国家自然科学基金项目(40671080)罗布泊“大耳朵”地区盐壳盐土特征及其发生学意义;土壤学自治区重点学科资助

摘　　要：本研究在野外综合考察基础上结合试验分析数据及遥感影像上提供的信息建立了罗布泊"大耳朵"干盐湖区盐壳分类体系,采用决策树方法对ETM+遥感影像进行分类,经过混淆矩阵检验,分类总精度达到86.3%,Kappa系数为0.8420。结果表明,罗布泊"大耳朵"干盐湖区不同盐壳类型在空间上呈环状分布与"耳轮"影像特征大致吻合;影像上同一条带的盐壳表现出相间分布的特点;相似形状的盐壳受地表湿度影响在遥感影像上呈现不同色调。It is of great significance to discover the salt crust characteristics and search for the scientific basis of the causes of ＂Great Ear＂ rings on remote sensing images in research of ancient environment evolution in Lop Nur ＂Great Ear＂ dry salt lake area. This study extracts the wetness index and part of texture characteristics from remote sensing images Landsat ETM＋ and establishes the digital elevation model based on image characteristics and DGPS data in the area. In the statistical analysis of the various characteristics of the numerical classification, select the best combination of bands on the basis of amendments to the optimum index factor EOIF. Combined analysis of experimental data and a comprehensive field study established the salt crust classification of the dry salt lake area, classified the remote sensing images ETM＋ by using decision tree technology, the overall classification accuracy comes to 86.3% and the Kappa coefficient is 0. 8420. Classification results show that different types of salt crust in space were showed ring-shape as same as the ＂Great Ear＂ rings of remote sensing images ; salt crust put up to intersection in the same band of images; similar shape of salt crust shows different hues by the impact of humidity of surface on the images.

关 键 词：罗布泊 “大耳朵” 决策树 盐壳分类 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]